ADVANCED CFD SIMULATION OF A COMPRESSED AIR INJECTION MODULE

Conference on Modelling Fluid Flow (CMFF 6) The 13th International Conference on Fluid Flow Technologies Budapest, Hungary, September 6-9, 26 ADVANCED CFD SIMULATION OF A COMPRESSED AIR INJECTION MODULE Huba NÉMETH 1, Gergely KRISTÓF 2, Viktor SZENTE 3, László PALKOVICS 4 1 Corresponding Author. Knorr-Bremse R&D Center Budapest, H-1119 Budapest, Major u. 69., Hungary. Tel.: +36 1 3829-828, Fax: +36 1 3829-81, E-mail: huba.nemeth@knorr-bremse.com 2 Department of Fluid Mechanics, Budapest University of Technology and Economics. E-mail: kristof@ara.bme.hu 3 Department of Fluid Mechanics, Budapest University of Technology and Economics. E-mail: szente@ara.bme.hu 4 Knorr-Bremse R&D Center Budapest. E-mail: laszlo.palkovics@knorr-bremse.com ABSTRACT Turbocharged engines exhibit the so called turbo lag phenomenon causing a disadvantageous response comparing to atmospheric or mechanically supercharged engines. There are many solutions developed to reduce the lag effect although each of them has some disadvantages. This paper investigates a booster module that injects compressed air to increase the response of turbocharged compression ignition engines of commercial vehicles. The investigations applied 3D computational fluid dynamic (CFD) simulations in stationary and transient manner as well. The transient simulations were applied as standalone and coupled calculations to a 1D engine simulation code to capture more accurate boundary conditions (BC). The drawbacks and problems of this new and advanced approach are emphasized in the paper. Due to the deformation of the flow field boundary surface during the dynamic process a deforming mesh method has been applied here. Instant shape of flow field boundary has been obtained by solving the equation of motion of the actuator rotor. By applying the above methods it was possible to validate the booster module design variants and to identify the main characteristics of the complete system including the engine. Keywords: CFD, compressed air, compression ignition engine, coupled simulation, moving mesh, turbo lag NOMENCLATURE I [A] current L [H] inductance F [N] force M [Nm] torque R [Ohm] electric resistance T c [Nm/A] torque constant U [V] voltage c [Nm/rad] spring coefficient k [Nms/rad] damping coefficient q [kg/s] mass flow rate p [Pa] absolute pressure φ [rad] angular position ω [rad/s] angular velocity Θ [kg/m 2 ] inertia Subscripts and Superscripts inlet at the inlet of the test section outlet at the outlet of the test section thr throttle valve sup supply valve p caused by pressure 1. INTRODUCTION Downsizing of a compression-ignition engine is an effective means for commercial vehicles for reducing fuel consumption but it requires supercharging to restore the full load torque. A compact and cost effective solution is with an exhaust turbocharger, which however has the drawback of a slow response to fast load demands by the driver. When the engine is increasing in speed, a delay occurs between advancing the accelerator pedal (which lets more fuel into the engine) and the time the exhaust pressure reaches a steady rate sufficient for the turbine. This delay is known as turbo lag. To reduce the turbo lag there are several methods such as application of downsized turbines with waste gate valves [1, 2], variable geometry turbo chargers [3] and recently the usage of turbo compounding by using an electric motor [4]. An attractive alternative here is using the external compressed air to boost the turbo charger dynamics since this medium is available on each commercial vehicle [5]. 388

The compressed air injection must occur in a short period to increase the charge in the cylinders of the engine making an increased fuel injection and thus an increased engine torque possible. Meanwhile the turbo charger is accelerated by a high margin by separating the turbo compressor output from the engine input for this short interval. This paper focuses on the CFD simulation of the characteristics of an air injection module for boosting the turbocharger dynamics. The first part is devoted to identify the static behaviour of the module and the second part investigates the complex transient dynamics of switching from the turbo compressor supply to the air injection and back again after reaching the desired response of the engine. The transient CFD simulation has been carried out using a 3D dynamic moving mesh governed by the actuator dynamics where the actuator dynamics has been coupled by separate ordinary differential equations. The transient simulation required special treatment of the boundary conditions to achieve a realistic behaviour of the module as being coupled to the compression ignition engine. For this purpose a 1D CFD engine simulation code of GT-Power [6] has been coupled to the 3D CFD code of FLUENT [7]. The 1D-3D coupling method is a new approach so there are just some few examples on its application [8]. The presented methods made it possible to validate the booster module design and identify the most important characteristics of the system. 2. SYSTEM DESCRIPTION The compressed air injection module is located in the high pressure intake part of the engine between the intercooler and the intake manifold (see Figure 1). The charger air coming from the intercooler is connected by the piping (2) to the air injection module (1) through the port (11), which conducts the air to the intake manifold by the piping (3) through the port (21). The external compressed air injection is provided by the compressed air system through the piping (5) on the port (12). The air injection module includes an electric controller which is supplied by the electric line (4). Intercooler 2 11 Figure 1. System layout 4 1 5 12 21 3 Intake manifold Additionally the air injection module has a throttle valve that can separate the port (11) form the ports (12) and (21). The operation of the system is as follows. Under normal conditions when the turbocharger runs at high speed the air is compressed by the compressor of the turbocharger that is cooled down by the intercooler. In this case the throttle valve of the air injection module is fully open and the air coming from the intercooler is conducted to the intake manifold. In case of low engine speeds when the turbocharger is unable to build up a significant charger pressure the engine is supplied by external compressed air injection from the compressed air system of the vehicle by a supply valve on port (12), meanwhile the throttle valve is fully closed to avoid a back-stream towards the turbo compressor. After reaching the desired charger pressure the supply valve is closed so the air injection is shut down and with a correct synchronization the throttle valve is opened and the normal air supply from the turbo compressor is restored and the engine can accelerate further with maximal boost. Switching from the turbocharger to external compressed air injection and back again during a highly transient acceleration raises a complex fluid dynamic process inside the air injection module that has to be strongly considered during the design. In order that the controller of the system is able to monitor these processes two pressure sensors are installed on the two sides of the throttle valve of the air injection module. Actuators operating the two alternative air inlets of the module have a significance that can strongly influence the performance of the module. Since the switching between the two air sources of the engine has to be executed in a short time to have a smooth operation both actuators should have high potentials. For this purpose the supply valve is designed as a diaphragm construction that is able to open and close big cross sections and can be operated in very short time due to low inertias. The control chamber pressure of the diaphragm is operated by a solenoid magnet valve. For the electric throttle valve operation a directly driven alternative has been selected, where the actuator is a so called DC torque motor. This is developed specially for throttle valves. The DC torque motor has no brushes so its angular operation range is limited. Since there is no transmission between the throttle valve axis and the torque motor the inertia of the whole actuator is very low and this allows a high response. For safety reasons the throttle valve is equipped with a return spring that can provide a well defined position even if the electric control of the system is lost. 389

3. STATIONARY OPERATION As discussed before the system has two distinct and important stationary operation modes. The one is the passive operation mode (throttle valve is open) when the engine is supplied by the turbo compressor with compressed air and the other is the active operation mode (throttle valve is closed) when the engine is supplied from the compressed air system of the vehicle by the air injection. These two operation modes well characterize the system and many important operation properties can be captured from the stationary results. The most important questions to be answered by stationary CFD simulations are as follows for the passive mode: - What is the effect of the open throttle valve to the charge of the particular cylinders of the engine? - What is the pressure loss caused by the throttle valve? - What about the velocity map and streamlines around the throttle valve? - Is the throttle valve stable in its open position if the return spring is broken? For the active operation mode the following questions are raised for the static investigations: - What is the throughput of the diaphragm supply valve? - Where should be the best location for pressure sensors to measure the static pressure on both sides of the throttle valve? - What is the maximal air speed during the air injection, where is it located? - How uneven can be the charging process of the particular cylinders of the engine due to the air injection? 3.1 Geometric Versions During the early development phase there were two main geometric versions of the module. The first one had a direct connection to the intake manifold of the engine by a flange interface (see Figure 2). The second version had circular interfaces on both high cross section ends fitted to rubber pipes (see Figure 3). The air injection module is represented by the transparent bodies in the figures. Since there is no external air injection in the passive mode the complex geometry of the compressed air supply valve head was removed and the air injection tube has only been retained on the geometry. Figure 3. Piping interface version for passive mode For the simulations of the active operation mode the geometry is basically the same with the exception that the input pipe is cut at the closed throttle valve and the compressed air supply valve head geometry is retained (see Figure 4). Figure 4. Supply valve head for active operation mode For all stationary simulations the intake manifold has always been included into the model geometry. Figure 2. Flange geometry version for passive mode 3.2 Numeric mesh Stationary simulations for passive mode have been prepared using the mesh represented in Figure 5. It can be seen that the mesh has a much finer structure in the vicinity of the throttle valve, with a total cell count of approx. 15. The linear cell size at and near the throttle valve is 2 mm. 39

3.4 Results Some typical results of the stationary models are seen in Figures 7 to 1 for both operation cases. The pressure drop along the analysed flow field can be calculated from the results of open throttle valve. The value of the pressure drop had been 7772 Pa. In case of closed throttle valve the boundary conditions had been set up differently, therefore from those results the mass flow can be calculated, which is.181 kg/s. Figure 5. Numeric mesh for passive mode In case of active mode in Figure 6, the mesh of the supply valve had to be refined similarly, again with a size of 2 mm, although on the diaphragm and in the opening the cell sizes had been reduced to 1 mm. Figure 7. Pressure in passive mode Figure 6. Numeric mesh for active mode 3.3 Physics and Boundary Conditions Stationary simulations with open throttle valve have been prepared with a constant-density gas model as the value of the Mach number had been quite low. The exact value of 3,187 kg/m 3 gas density had been calculated from the 3 bar inlet pressure and 55 C inlet temperature. Stationary simulations with closed throttle valve have been prepared with an ideal-gas model. This calculates the density from the state equation of ideal gases. The calculations of the turbulent transport values had been carried out using the Renormalisation Group (RNG) k-ε turbulence model. The boundary conditions of the turbulence models had been set using the hydraulic equivalent diameter of the inlet and outlet pipe sections, assuming 1% turbulence intensity. These values are not expected to affect the simulation results significantly. The boundary conditions for the two stationary cases are shown in Table 1. Table 1. Stationary boundary conditions Case Outlet Compr. inlet Inlet Passive.25 kg/s closed 3 bar Active 3 bar 7.5 bar closed Figure 8. Velocity in passive mode Figure 9. Pressure in active mode 391

Figure 1. Velocity in active mode In passive operation mode the open throttle valve has caused almost no perceptible disturbance on the main stream. The maximal velocities are in a low range below 3 m/s even at full engine load and nominal speed. The throttle valve can stay stable in its open position even with broken return spring. At active operation the maximal air speed reaches the sonic speed limit. The charge process of the particular cylinders are less disturbed in case the air injection is directed to the opposite pipe wall than addressed directly into the intake manifold. switched to the other state then the movement of the diaphragm is done in a very short time of about 1 ms. So the diaphragm dynamics was considered as a binary on/off valve with a predefined actuation delay caused by the magnet valve of 1 ms. In case of the throttle valve actuation the situation is different it has a slower dynamics compared to the supply valve, moreover the air stream and pressure conditions have a bigger influence on its operations. The throttle valve axis has an offset in order that the air stream can open it in case of emergency when the return spring is broken. Due to the above reasons the throttle valve with its actuator is considered to be included in the calculations. 4.1 Handling of deforming mesh Fluid mechanical simulation using deforming meshes is a new and innovative option which has been provided by the new FLUENT 6.x version. Currently some compromises had been necessary though. It is particularly tough to deal with small gaps, especially when there is a significant shear in the mesh deformation. Therefore the original geometry had been fitted with a spherical-shellshaped ring as it can be seen in Figure 12. This ring had been filled with cells of very high porosity. Therefore the cells can deform freely allowing the deforming mesh function to work appropriately, while at the same time the flow is heavily constrained in this region. For a fully closed throttle valve it means that only 2.6% of the base flow (stationary flow with open throttle valve) had been able to escape through this porous region. Tetrahedral elements have been applied in the dynamically meshed zone. Local mesh reconstruction was triggered by exceeding values of equiangular skew. For the transient operations the intake manifold is not more included into the geometry domain. Figure 11. Surface pressure in active mode The static surface pressure distribution (see Figure 11) shows an even pressure on a big surface providing a free pressure sensor positioning. 4. STANDALONE TRANSIENT OPERATION The transient simulations address to investigate the time domain effects of the system. According to the preliminary investigations by using lumped parameter models on the two actuators of the system one could conclude that the supply valve diaphragm axle has a very fast dynamics itself although the complete air injection actuation dynamics is mainly influenced by the response of the actuating magnet valve. If the magnet valve has Figure 12. Porous zone for the deforming mesh The time step size of.1 s for the deforming-mesh simulation had been optimised using preliminary test runs for the whole time domain. Mesh independency of the simulation results could not be analysed due to limited timeframe of the project. The standalone simulations applied the constant-density gas model. 392

4.2 Boundary conditions for unsteady flow model For standalone simulations the boundary conditions had been modelled as time-independent characteristic curves. Since the objective of these calculations had been the simulation of the closing and the opening transient, which is a quite small time domain, the boundary conditions can be described using static characteristics. These characteristics describe the inlet total and outlet static pressure as a function of the mass flow as it can be seen in Figures 13 and 14. The resulted constant pressure distributions have been applied to the inlet and outlet sections of the booster module. pressure[pa] 1.6 x 15 1.55 1.5 1.45 1.4 1.35 1.3 1.25 1.2 1.15 Inlet BC - Turbo compressor side Outlet BC - Engine side Operation point movement during throttle closing 1.1.2.4.6.8.1.12.14 mass flow rate [kg/s] Figure 13. Time-independent boundary conditions for throttle closing 2.9 3 x 15 Inlet BC - Turbo compressor side Outlet BC - Engine side di U R Tc = I ω, (1) dt L L L dω M p + Tc I kω M f sgn( ω) cφ = dt Θ (2) dφ = ω dt. (3) The torque M p caused by the static pressure distribution on the throttle plate is fed back from the surface torque integral from the CFD domain. On the other hand the throttle plate position defines the geometric domain that is given by the angular position from Eq. (3). Eq. (1) to (3) have been integrated by second order Runge-Kutta scheme implemented in C-language under Fluent. 4.4 Results The throttle dynamics obtained from the simulation is depicted in Figure 15. Duty [-] Voltage [V] Speed [rpm] 1.5 -.5 2-2 5 Current [A] Position [deg] -5 5-5 5 Time [s] 2.8 Figure 15. Throttle valve dynamics pressure[pa] 2.7 2.6 2.5 Operation point movement during throttle opening The velocity maps as function of the resulted throttle valve positions for open, 3 degrees, 6 degrees and closed states are given in Figures 16 to 19. 2.4 2.3.2.4.6.8.1.12.14.16.18 mass flow rate [kg/s] Figure 14. Time-independent boundary conditions for throttle opening 4.3 Dynamic model of the throttle valve Based on electrical and mechanical state balances the throttle valve dynamics has been calculated using the ordinary differential equations based on first engineering principles for conservation of magnetic linkage and mechanic impulse. Figure 16. Velocity for open throttle valve 393

5.1 Engine model The boundary conditions had been provided by GT-Power in such a way that while Fluent s inputs are the mass flow rates at both boundaries, GT- Power receives pressures as inputs on the connecting interfaces. The high pressure intake portion of the GT-Power engine model is shown in Figure 2. On left side the 3D Fluent domain called module with its CFD interfaces, on the right side the intake manifold with the cylinders are seen. Figure 17. Velocity for 3 degrees position Figure 2. Portion of GT-Power model with coupled 3D domain Figure 18. Velocity for 6 degrees position The results have shown that the throttle valve dynamics is only slightly influenced by the pressure distribution on the throttle plate under typical operation conditions. Figure 19. Velocity for closed throttle valve 5. COUPLED TRANSIENT OPERATION In coupled simulations everything but the gas density and the boundary conditions had been the same as in standalone cases. As the coupling proved to work correctly only with compressible gases, therefore the ideal-gas model must have been used. In order to have realistic operation conditions the GT-Power engine model has been coupled to a GT-Drive dynamic vehicle model. As test case for the simulations a step response situation has been used by emulating a kick down acceleration on an uphill slope in 4 th gear (of a 6 gear gearbox) with an initial engine speed of 12 rpm. In contrast to conventional GT-Power models with turbocharger it had a big importance of modelling the turbo compressor appropriately since it operates in the surge region during the air injection actuation for some engine cycles. However it has a crucial importance on the resulted boundary condition of the module inlet. For this purpose a mean value compressor surge characteristics has been used based on preliminary dynamometer measurements. 5.2 Results The coupled iteration time steps are driven by GT-Power that is followed by Fluent however at a particular time step there is no coupled iteration between the two codes in order that the solution could converge at each time step. Due to this reason the coupled convergence is not guaranteed even under highly relaxed conditions. In conclusion the coupled simulations required extensive trials to reach certain convergence properties. The compressor map with the operation points is shown in Figure 21 with normal and boosted acceleration cases. One can see that the compressor operates for some engine cycles in the surge area if the compressed air injection has been applied. However the mean mass flow rate did not reduce to 394

zero during this phase. Since the mean power consumption decreases strongly during the boost actuation the charger speed increases very fast. Figure 21. Compressor map and operation trajectory The engine related inlet and outlet pressures of the module are shown in Figure 22. The accelerator pedal has been kicked down at 1 s. The duration of the applied air injection was.4 s long after kick down. Figure 22. Inlet and outlet pressures In the normal acceleration case without boost the pressure build up time results in 4.5 s due to the low initial engine speed and the stiff characteristics of the load torque working in high gear and uphill. In the boosted acceleration case the booster outlet (intake manifold) pressure raises fast due to the applied air injection and the pressure build up time results in.2 s. In the meantime the booster inlet (intercooler) pressure approaches the outlet pressure. This boosting effect has been successfully tested and verified at extensive dynamometer and vehicle test as well. The coupled transient calculations have confirmed that the air stream is able to open the closed throttle valve if the return spring is broken. dynamics of turbocharged compression ignition engines. The simulations have covered stationary and transient cases that made it possible to capture the most important properties of the system. The transient simulations applied dynamic moving mesh techniques to involve the relevant actuator dynamics that were described by ordinary differential equations. In a second step the transient simulations have been coupled to a 1D engine simulation code to provide more accurate boundary conditions. By applying the above methods it was possible to validate the booster module design variants and to identify the main characteristics of the complete system including the engine. REFERENCES [1] F. M. Schittler,V. Rubi, G. Fraenkle., 1993, Mercedes-Benz Commercial Vehicle Engines For Euro II And EPA 94 Emission Limits, SAE 931636, pp. 1-16. [2] B. Ubanwa, M. Kowalczyk., 1993, Optimization Procedure for Matching Exhaust Waste-Gate Turbocharger to Automotive Diesel Engine, SAE93112, pp. 257-275 [3] Y. Chi, J. Cheong, C. Kim and K. Choi, 22, Effects of VGT and Injection Parameters on Performance of HSDI Diesel Engine with Common Rail FIE System, SAE 22-1- 54, pp. 1-7 [4] U. Hopmann and M. C. Algrain, 23, Diesel Engine Electric Turbo Compound Technology, SAE 23-1-2294, pp. 1-7. [5] H. Németh and E. Gerum, 25, Einrichtung zur Frischluftversorgung einer turboaufgeladenen Kolbenbrennkraftmachine und Verfahren zum Betrieb derselben, Deutsche Patentanmeldung Nr. 12551687.4 (patent pending) [6] Gamma Technologies, 24, GT-Power User s Manual, v6.1, pp. 1-286. [7] Fluent Inc., 25, Fluent User s Manual, v6.2, pp. 1-473. [8] U. G. Riegler and M. Bargende, 22, Direct Coupled 1D/3D-CFD-Computation (GT- Power/Star-CD) of the Flow in the Switch-Over Intake System of an 8-Cylinder SI Engine with External Exhaust Gas Recirculation, SAE 22-1-91, pp. 1-12. 6. SUMMARY The paper describes a 3D CFD investigation of an air injection module used to improve the 395